The Agent Skills Directory

[SAFE]: No high-risk security patterns were identified. The skill is designed to restore project context using localized scripts and restricted agents.
[COMMAND_EXECUTION]: Executes local bash scripts for project discovery and validation. These scripts use safe shell patterns, such as fixed-string grep and path validation, to mitigate injection risks. The validate.sh script explicitly prevents path traversal attacks by checking for .. and verifying the path prefix.
[PROMPT_INJECTION]: The skill reads project files (checkpoints and PRDs). This is a potential indirect prompt injection surface; however, the risk is mitigated by using a dedicated sub-agent with a limited toolset (Read, Glob, Grep, LS) and requiring user confirmation before taking actions based on the parsed data.

dev-resume